A significant portion of microbial genomes remain uncultured, and many proteins lack functional annotations. This research introduces LookingGlass, a deep learning model that creates contextually aware representations of short DNA reads, distinguishing reads based on function, homology, and origin. LookingGlass, through transfer learning, successfully identifies novel oxidoreductases, predicts enzyme optimal temperatures, and recognizes DNA sequence reading frames. This model provides functionally relevant representations of unknown sequences, shedding light on 'microbial dark matter'.
Publisher
Nature Communications
Published On
Nov 16, 2022
Authors
A. Hoarfrost, A. Aptekmann, G. Farfañuk, Y. Bromberg
Tags
deep learning
microbial genomes
oxidoreductases
transfer learning
functional annotations
DNA sequences
enzyme prediction
Related Publications
Explore these studies to deepen your understanding of the subject.
